公开(公告)号：MY173744A

公开(公告)日：2020-02-18

申请号：MYPI20092455

申请日：2009-06-15

Applicant: MIMOS BERHAD

Inventor： LIANG KIM MENG ,  ZULAIKHA KADIM ,  AHMED BAHAA AL-DEEN ,  LIM MEI KUAN ,  TANG SZE LING

Abstract: A method to determine point-of-sight from a sequence of image is described with the following steps. First, visual fixation of human is determined from a sequence of image (20). Then, gaze intersection of visual fixation is detected (40). Multiscale weight is then computed (61) which assign weightage to each pixel which reflect the significance of a particular point, assuming that a point nearer to the detected intersection has a higher possibility of being point-of-sight. Every weightage map is next superimposed and new weightage point for every coordinate is accumulated (62). Later, the master point-of-sight map is updated (63) and harmonized (64). A high value of weightage indicates high probability of point-of-sight. Finally, the weightage of each pixel in the master point-of-sight map is analyzed and compared against predefined rules to indicate the level of suspiciousness in the scene and actuate alert accordingly 80.